[jetty-users] JDBCSessionManager single point of failure

[Roger Armstrong <roger@xxxxxxxxxxxxxxxxxxxxxxx>]

We using embedded Jetty 7 for our Wicket apps with the JDBCSessionManager for clustering behind HAProxy with sticky sessions. Works great, but if the JDBCSessionManager MySQL database becomes unavailable for any reason, requests to the cluster just hang. What we'd really like would be that Jetty detects that the database is unavailable and keeps on serving requests without replicating the session (and when the database becomes available again, starts replicating again). After all, the goal of clustering with JDBCSessionManager is to improve uptime - having the JDBCSessionManager database as a single point of failure for the whole cluster doesn't seem conducive to this goal.

I understand that this only makes sense in a sticky sessions load-balancer configuration (since otherwise the load-balancer might send a request to a server without the replicated sessions), but for stateful web-applications, sticky-sessons are anyway the most meaningful scenario.

Is there any way to achieve this in the current implementation of the JDBCSessionManager?

By the way, we tried various other clustering strategies such as Terracotta, but for various reasons (mostly simplicity and the reliability which ensues from simplicity) we reverted to the JDBCSessionManager.

regards,

Roger Armstrong